MARKING AND CLITIC DOUBLING IN ROMANIAN

(1)

RRL, LXII, 4, p. 393–409, 2017

MARKING AND CLITIC DOUBLING IN ROMANIAN

VIRGINIA HILL¹, ALEXANDRU MARDALE²

Abstract. The key question in this paper is the following: why do Differential Object Marking (DOM) and Clitic Doubling (CD) interact in Modern Romanian, since that was not necessarily the case in Old Romanian? The hypothesis we defend relies on the presence of a topic feature at the left periphery of DOM-ed noun projections: the bleaching of this feature, reflected through the grammaticalization of the DOM particle pe, triggers changes in the implementation of feature checking; in particular, it resort to CD as a means of supplementing the checking function of pe. The corollary of this analysis is that the emergence of the CD/DOM interaction depends on a major parametric shift, whereby Clitic Left Dislocation (CLLD) is generalized in the language to the detriment of topicalization; CD is a sub-case of CLLD. Empirical evidence comes from a corpus of original and translated texts from the 16^th century.

Keywords: Differential Object Marking, Clitic Doubling, Clitic Left Dislocation, grammaticalization, reanalysis, preposition, (topic) marker, Case, (Old) Romanian.

1. INTRODUCTION

Differential Object Marking (DOM) with pe in Modern Romanian (MR), is tightly related to Clitic Doubling (CD): in (1), both CD and DOM are obligatory.

(1) Am strigat-*(oj) pe eaj. have called-her DOM her

‘I called her.’

Formal approaches to constructions as in (1) generally use CD to justify DOM: the clitic absorbs the Accusative Case of the verb, so the direct object needs a prepositional Case assigner (Kayne 1975, Dobrovie-Sorin 1990). There are, however, studies that bring evidence against this analysis, cross-linguistically (Massey 1991, Suñer 1988) and for Romanian (Gierling 1997 a.o.). For Romanian, evidence against a Case approach comes from the possibility of having independent

1 University of New Brunswick – Saint John, [email protected]

2 INaLCO de Paris – SeDyL UMR 8202 CNRS, [email protected]

(2)

Virginia Hill, Alexandru Mardale 2 394

occurrences of CD and DOM in Old Romanian (OR), and also, the use of DOM without CD in MR when bare quantifiers are present (e.g., pe nimeni ‘nobody’).

In fact, the studies on OR reveal that DOM is, by default, dissociated from CD in texts, and neither CD nor DOM were obligatory with any class of nouns (Avram and Zafiu 2017, Heusinger and Onea Gáspár 2008, Hill and Tasmowski 2008, Mardale 2015, Pană Dindelegan 2014 a.o.). These studies explicitly relate pe-DOM in OR to discourse needs (i.e., for marking salient information or prominence), not to functional needs (i.e., for Case assignment).

Considering this background, the question we address in this paper is when and why the interaction between DOM and CD emerged in OR, in a way that makes both operations obligatory in the context of (1) in MR. We consider that the CD/DOM interaction is related to the downwards reanalysis of pe, from preposition (P) to nominal topic marker (K), and further to the switch in interpretation between contrastive topic for independent DOM to familiar topic for CD/DOM constructions. We relate these changes to the featural make-up: the valued contrastive topic feature becomes unvalued and is transferred down to the inflectional domain (i.e., to D(eterminer)). As pe has no access to D, an agreement relation between D and the clitic in T arises to implement feature checking and valuation. However, this alternative means for feature checking is possible only insofar as the clitic is available in the speaker’s grammar.

As sources of data, the OR examples provided in this paper come from translated and original texts and documents, mainly from the 16^th century. For statistics, we use two 16^th century texts: one written directly in Romanian (DÎ) and one translated (PO). In this way, we control for the genuine extent of CD/DOM in the Romanian grammar, as opposed to translation artifacts under the impact of a foreign grammar. However, when it comes to illustrations, we generally aim at syntactic minimal pairs, which may not be available in those two texts. For this purpose, we searched the entire corpus of 16^th century texts (especially Coresi’s prints).

2. DIACHRONIC PERSPECTIVE 2.1. About pe-DOM

Recent studies (Hill 2013, Mardale 2015) argue that, in selected contexts, the preposition pe was reanalyzed as a DOM particle, that is, from lexical preposition to a functional marker, being semantically bleached (i.e., its meaning of location, purpose and so on is lost) but functionally enriched with a discourse feature that triggers salience for reading. We label this feature as contrastive topic (as opposed to aboutness or familiar topic), and use it as an umbrella for emphasis, listings or

(3)

any contrast that does not imply exclusivity but produces foregrounding (see Lee 2003)³.

We refer the reader to Hill and Tamowski (2008) and Hill (2013) for arguments toward a prominence/contrastive topic analysis of DOM in OR. Here, we summarize the main points:

• Pe-DOM is in complementary distribution, in the same context, with non-DOM-ed direct objects, as in (2) and (3), the option depending on whether the direct object is or is not foregrounded.

• The foregrounding with pe-DOM arises from a contrastive topic (versus contrastive focus) effect because it allows for list readings, as in (3). That is, in (3a), DOM indicates that the listing and the presentation of the individuals is what counts for the information. On the other hand, in (3b), the important part of the information is what those individuals said, not who they were, so DOM does not apply.

(2) a. ascultaţi mine (PO, 73) listen me

‘listen to me.’

b. au ascultat pre mine (PO, 119) have listened DOM me

‘they have listened to me.’

(3) a. am întrebat pre toţ fraţii miei şî pre toate rudele mele have asked DOM all brothers-the my and DOM all relatives-the my şî pre toţ meg[i]eşîi di sat (DÎ, VI, 1579-80)

and DOM all landowners-the from village

‘I consulted all my brothers and all my relatives and all the landowners from the village.’

b. amǔ întrebat nepoţii mii şi ruda mea (DÎ, LXVI, 1586) have asked nephews-the my and relative-the my

‘I consulted my nephews and my relative.’

• Since the purpose of pe-DOM is a discourse effect, we expect it to apply irrespective of the semantic noun class, which is the case, as in (4). Although animates are preferred for DOM in texts, inanimate nouns are also DOM-ed in OR (but not in MR).

(4) a. a ceti pre acest letopisăţ mai mult (Neculce, 5) to read DOM this chronicle more much

‘to further read this chronicle.’

b. va săruta pre cinstitele ale lui mâini (Cod Tod, 85r) will kiss DOM honorable-the of his hands

‘he will kiss his honorable hands.’

3 In Lambrecht (1994: 97) Contrastive Topics provide clarification when several options are possible; for example, “I saw MARY yesterday. She says HELLO”. CTs also allow for listing readings, as in “I saw MARY and JOHN yesterday. SHE says HELLO, but HE’s still ANGRY at you”.

(4)

c. Şi deaderă lui Iacov pre bozii cei striini (BB, FacereaCapXXVIII) and gave DAT Jakob DOM weeds-the the foreign

‘And they gave to Jakob the foreign weeds.’

To summarize the situation in OR, in unselected contexts, the preposition pe is maintained with its lexical properties, while in selected contexts, it is semantically bleached and reanalyzed downwards, as a functional element that foregrounds the (in)animate direct object.

2.2. Clitic doubling (CD)

If pe-DOM is foregrounding, CD is backgrounding (see also Frascarelli and Hinterhölzl 2007 for pointing out that clitic pronouns are generally the items that qualify for their Top-familiar). OR texts attest to the use of CD independently of pe-DOM. The examples occur mostly in translated texts and with strong personal pronouns. In OR, strong pronouns in direct object position may occur with or without CD, as in (5a) versus (5b, c). However, in (5a), the strong pronoun may be interpreted as new information/presentational focus. In order to avoid such possibility, translators tend to CD the pronouns, and thus ensure a neutral reading as required in their original (Hill and Tasmowski 2008). That is, when the pronoun is under CD, the reading becomes obligatorily neutral: e.g., in (5b), the new information is the right dislocated părintele, not the in-situ mine.

(5) a. ascultaţi mine (PO, 73) hear me

‘hear me!’

b. m-au tremis mine părintele (CEV, 140) me-has sent me priest-the

‘The priest sent me.’

c. cum să te cunosc tine (PO, 292) that SUBJ you know you

‘so that I know you’

This use of CD is not a calque, but looks like a translation artifact (Hill and Tasmowski 2008). The examples in (5) come from translations from different languages (i.e., Church Slavonic for (5b) and Hungarian for (5a, c)), and CD applies in the same way. It is likely that CD was available in OR grammar, perhaps as an archaic property that faded by the 16^th century. The presence of CD without pe-DOM in Aromanian (Mišeska-Tomić 2006) suggests that this operation was present in OR before the dialectal split. Whatever the situation may have been, the independent use of CD appears unproductive in texts and fading from the language.

A reason for this unproductivity is suggested in section 4, after we formalize the CD/DOM interaction. At this time, the important point is that the earliest texts

(5)

also show some CD/DOM interaction, both in original and in translated texts, as shown in (6).

(6) a. te-au adus Domnul pre tine afară (PO, 221) you-has brought Lord-the DOM you outside

‘the Lord brought you out.’

b. ne-au făcutu pe noi mărţ de tot (DÎ, I, TR, 1595-1596) us-have done DOM us ko of all

‘they completely knocked us out.’

This interaction has a low incidence in 16^th century texts (e.g., 9.3% of DOM-ed DPs in DÎ, half of which display strong personal pronouns) but becomes productive in the following century.

Summarizing section 2, we can point out that the CD/DOM interaction emerges by mid-16^th century and becomes the default option by the end of the 18^th century. Up to that point, CD and pe-DOM could operate independently of each other, under discourse triggers: backgrounding for CD versus foregrounding for DOM.

3. THEORETICAL FRAMEWORK 3.1. DP left periphery for DOM

Recent studies in cartography and beyond argue for a similar organization of the left periphery in CPs and DPs (Aboh 2004, Aboh et al. 2010, Giusti 2006, Haegeman 2004, Szabolcsi 1994, Wiltschko 2014), that is, discourse features are associated with the highest (phase edge) functional head in both domains. We proceed along the lines in Giusti (2012), where the highest projection of a nominal phrase is K(ase)P, the functional equivalent of CP. Since we adopt this analysis, henceforth we refer to nominal phrases as KPs instead of DPs.

Accordingly, for OR DOM, K has the contrastive topic feature (in the way C is associated with Topic), and the downward reanalysis of pe means direct merge in K instead of P, as in (7).

(7) P-pe → [KP K-pe [DP D…]]

This analysis captures not only the discourse effect of pe-DOM in OR but also the intuition that pe is a Case marker: if K is equivalent to C, then it has two sets of features, namely, φ-features (involved in Case checking) and discourse features. As pe is the only spellout for K, it covers the entire featural make-up of K, even if Case checking proceeds independently of pe.

(6)

3.2. CD: Delfitto (2002)

For clitic constructions, Delfitto (2002) argues for a treatment in terms of unsaturated expressions: the argument position related to an (object) clitic is re- opened and interpreted as a variable bound by a λ-operator. Thus, sentences involving pronominal clitics are predicates represented as λ-abstracts. For example, a construction like It. lo legge ‘reads it’, where legge is a two place predicate, has the semantic representation: lo legge →λx λy (x legge y).

For Delfitto, lo does not saturate the object argument position, so legge maintains its argument structure unchanged, which predicts the co-occurrence of lo with a selected DP. However, in the presence of lo, the λ-abstract is partially encoded in syntax: clitic constructions are grammatical tools encoding λ- abstraction over the argument positions of (verbal) predicates.

Syntactically, the λ-abstract is associated with a Top head (the subject of predication) that attracts a KP to Spec, TopP, as its argument. Top is associated with a PRED feature, and [+PRED] attracts the KP to its Spec, yielding topicalization in some languages (e.g., English). However, [-PRED], as in Romance languages, cannot attract the KP and the λ-abstract is encoded instead as an Agr feature on T, which is spelled out as a clitic pronoun. The pronominal clitic activates the KP movement to Spec, TopP. Importantly, the clitic is not an argument of the λ-abstraction, but only a functional feature that mediates the syntactic mapping of the semantic relation through Clitic Left Dislocation (CLLD).

The implementation takes two forms: hidden CLLD when the KP is non-lexical, or overt CLLD when the KP is lexical. Obligatory movement applies to both types of KP.

Delfitto’s analysis is compatible with CD (a construction he does not discuss) in the sense that any clitic construction gives rise to unsaturated λ-abstracts, where there is a semantic subject of predication (saturating the λ-abstract in order to give rise to a proposition), independently of the specific syntactic execution of this subject of predication. In this paper, we consider CD as a sub-case of CLLD, when no movement applies.

3.3. CD/DOM: Miyagawa (2010)

Delfitto’s definition of the pronominal clitic as an Agr element in a topic chain naturally sends us to Miyagawa (2010) who also argues, independently of clitics and outside the Romance philum, that [topic]/[focus] features must be treated as a type of discourse Agr at C.

In Miyagawa (2010), syntactic agreement is justified as a way of establishing a functional relation, such as subject-predicate or focus-presupposition.This directly subsumes the mapping of λ-abstracts in Delfitto (2002), also concerned with argument-predicate relations. Thus, C has two sets of features concerned with Agreement: phi-features (φ) and discourse features (δ). In line with the hypothesis

(7)

on feature transfer from C to T (Chomsky 2008 et seq), it follows that languages differ insofar as the feature transfer may apply to both feature sets, or only to one or to none of them. The possible range of variation is illustrated in (8):

(8) Miyagawa’s 2010 typology:

Category I: Cφ, Tδ Japanese Category II: Cδ, Tφ English Category III: C, T φ/δ Spanish Category IV: Cφ/δ, T Dinka

Within this framework, the generation of a clitic object in Romance languages signals the transfer of Agr-δ to T, as predicted for Category III in (8).

Importantly, while TopP is projected within CP, the agreement it requires for a functional relation with the predicate is transferred to T and spelled out as a pronominal clitic. Essentially, we must distinguish between two types of discourse features: one that is concerned with the functional relation of Top/Foc and the predicate (i.e., δ = discourse Agr); and one that maps the expressivity (Miyagawa’s term) inherent to the discourse feature (e.g., aboutness, familiarity or contrast). The expressive feature is always mapped to C (i.e., TopP/FocP are in the CP field), whereas the Agr-δ feature is subject to transfer, on a par with the Agr-φ set at C (Chomsky 2008 et seq).

In conclusion, we adopt Delfitto’s view of the clitic as the spellout of an Agr functional relation with no involvement in the argument structure of the verb, but redefine it in Miyagawa’s framework as an Agr-δ item subject to transfer from C to T.

4. ANALYSIS

In section 4.1, we assume previous semantic accounts of DOM in Romanian with no further discussion. In section 4.2, the focus is on how syntax contributes to those readings, mainly through changes that affected the featural make-up of the relevant syntactic structures.

First, we remind the reader that in the 16^th century texts, a nominal direct object, for example a strong personal pronoun, may appear with no marking at all (9a), with CD only (9b), with DOM only (9c), or with both CD and DOM (9d). The examples come from the same text, indicating intra-speaker variation.

(9) a. rugăm tine ca drag părintele nostru (PO, 9) NO MARK implore you as beloved parent-the our

‘we implore you, as out beloved parent.’

b. te cunosc tine (PO, 292) CD

you know you ‘I know you.’

(8)

c. cel putearnic va blagoslovi pre tine (PO, 175) DOM the powerful will bless DOM you

‘the powerful one will bless you.’

d. te voiu aduce pre tine de acolo (PO, 162) CD/DOM you will bring DOM you from there

‘I will bring you from there.’

Intra-speaker variation signals that the option for one or another construction must be motivated by changes in the interpretation, since free variation is generally avoided in the language. The hypothesis regarding the merging of a contrastive topic feature in K conforms to this prediction.

4.1. Referential stability

The first question arising from (9) concerns the loss of the option in (9a).

This is related in the literature to the noun semantics: animate and definite nouns are DOM-ed, while inanimate and indefinite nouns are not (Bossong 1991, Comrie 1989 a.o.). This approach is not always adequate for Romanian: see (10), where inanimates undergo CD/DOM, and (11), where animates with pure role reading do not undergo CD/DOM (Gierling 1997), against predictions.

(10) Dintre toate cărţile ei am ales-*(o pe) aceasta.

from all books-the her have chosen-it DOM this

‘From all her books I chose this one.’

(11) Am pupat mireasa. // Am pupat-o pe mireasă.

have kissed bride-the have kissed-her DOM bride

‘I kissed the bride (whoever she is)’ // ‘I kissed the bride (a specific one).’

In response to examples as in (10) and (11), the concept of reference was added to the semantic properties of animacy and definiteness: DOM ensures referential persistence in the discourse (Chiriacescu and von Heusinger 2010).

While this may be statistically true for a given text, the fact remains that MR displays not only DOM but also simultaneous CD, and that CD/DOM occurs in out-of-the-blue utterances to which referential persistence does not apply.

However, it is uncontroversial that human referentiality is pervasive with DOM in MR. Farkas and von Heusinger (2003) establish a scale of referential stability, where items with the highest score of referentiality require DOM, whereas indefinites do not, unless the referential content is somehow recovered from the syntactic context (Tigău 2015). In other words, DOM is triggered by the deixis potential of a human and prominent direct object, which is, for example, very high with personal pronouns as in (9).

(9)

Basically, the studies on the semantics of DOM constructions argue that the preponderance of animate nouns for DOM arises from the natural potential of animate nouns to have higher referential stability and trigger discourse prominence.

The point we raise here is that this semantic/pragmatic property can be mapped to syntax just by DOM, with no need of CD, as seen in various languages (Comrie 1989). So the question is why is CD involved, in addition to DOM, in MR, but not so much in OR?

4.2. Syntactic properties

The first syntactic observation is that CD concerns the CP/TP relation, whereas DOM concerns the internal structure of the KP, and that these domains interact: the clitic used for CD obligatorily agrees in φ-features with the KP-pe (instead of displaying some invariable form), and the fact that the agreement relation crosses pe (supposedly a preposition) needs explaining. We do that by relating the checking of C [top] and the checking of K [top] via the clitic.

For DOM without CD in OR, we assume the reanalysis in (7), where pe is directly merged in K (instead of P) to check its uninterpretable but valued contrastive topic feature (we follow Pesetsky and Torrego 2007 for separating interpretability and valuation of features).

Evidence for this reanalysis comes from (12), showing that the lexical preposition pe in OR could select a coordination phrase with two KPs (12a), whereas that kind of coordination is not possible in selected contexts, with pe-DOM (12b); for the latter, pe is obligatorily repeated, indicating that it is inside the KP, not the selector of it, hence, lower in the structure. The examples contain strong pronouns, which are obligatorily DPs (vs. NPs), and a possessive adjective, which also indicates at least the DP (vs. NP) level of the nominal projection in (12a). This distinction is important because coordination at the NP level is allowed under pe-DOM (e.g., I-am invitat pe Dan şi Marius ‘I invited Dan and Marius’), unlike the KP/DP level.

(12) a. era mâniiat pre [slugile sale şi mine] (PO, 140) was furious DOM servants.the his and me

‘he was furious with his servants and me’

b. şi-i învaţă să facă cum se cade, [şi pre ei şi pre noi] (CEV, 90) and-them taught to do how is fit and DOM them and DOM us

‘and he taught them to do what is befitting them and us’

This amounts to Category II in (8), as [δ] remains at K, while [φ] is transferred, as in (13).

(13) +DOM/-CD

[KP K-pe [utop/δ], [φ] [Dmax D[φ]….]] → pe checks [utop/δ]

(10)

Merging pe in K implements the checking needs, while valuation is inherently provided for [top]. This process is independent of the contrastive feature of Top in C, which has wide scope and does not interfere with DOM.

On the other hand, when CD interacts with DOM, the topic feature of C and K display some kind of dependency, since K loses the ability to locally associate with a contrastive topic reading, and in fact, KP can only have a familiar topic reading as long as it stays in situ. If the DOM-ed KP is fronted, it gets an aboutness reading. We account for this peculiarity by assuming featural bleaching: the topic feature of K becomes underspecified, instead of entering the derivation with the contrastive value. In this case, pe can check the topic feature, as in (13), but cannot value it. Valuation needs to come from another source or the derivation crashes.

CD provides the alternative source, when an agreement relation is established between D and the clitic in T. For this agreement to happen, D must have not only [φ] but also [δ], in order to match the features of the clitic and copy its values.

Considering (8), this amounts to a typological switch from Category II to Category III, within KP, as in (14).

(14) +DOM/+CD

[KP Kpe [utop, δ], [φ] [DP D[δ],[φ]….]] → pe checks K[utop]; clitic checks/values D[δ],[φ]

The weakening of the topic feature on K goes hand in hand with an increase in the mapping of animacy and in the deixis potential of the DOM-ed nominal.

Both animacy and deixis involve the [φ] set of D: animacy is mapped as inner aspect on the N root, on a par with or in complementarity with the [number] feature (Wiltschko 2014); while the deictic property pairs with the need of D for valuation of its [person] feature (Ritter 1995). These properties indicate that the [φ] set also underwent bleaching and depends on valuation from the clitic.

There is empirical support for the contrast between (13) and (14), coming from statistics and from fronting tests. First, statistically, the texts indicate that the CD/DOM interaction spread at a higher rate after CLLD (versus topicalization) was established as the only way of fronting constituents to the left periphery of clauses. The competition is illustrated in (15).

(15) a. vor vedea tine eghipteanii (PO, 44) - CLLD will see you Egyptians.the

‘the Egyptians will see you.’

b. mare oamet te voiu face (PO, 287) hidden CLLD

great man you will make ‘I’ll make you into a great man.’

c. cum pre mine încoace m-aţi vândut (PO, 159) overt CLLD as DOM me here me-have sold

‘as you sold me here.’

d. mine vor omorî şi tine vor ţinea (PO, 44) topicalization me will kill and you will hold

‘me, they’ll kill, and you, they’ll hold.’

(11)

Table 1 reports the findings on CD and DOM options in two 16^th century texts, one a translation, the other a document written directly in Romanian.

Table 1

Distribution of CD and DOM in OR

PO DÎ

pers. pronouns

+/- clitics; 829

other DPs +DOM total: 312

pers. pronouns +/-clitics; 992

other DPs +DOM total: 278

hidden CLLD 48,48% N/A 97,4% N/A

overt CLLD 2,41% 4,49% 0% 6,48%

+CD/-DOM 2,53% 0% 0% 0,72%

-CD/+DOM 21,83% 81,7% 0,81% 78%

+CD/+DOM 8,20% 2,56% 1,52% 4,32%

-CD/-DOM 11,82% default option 0% default option

Topicalization 4,70% 11,22% 0,2% 10,43%

For Table 1, we proceeded to a comprehensive search of direct objects in the two documents. The personal pronouns surveyed cover both clitic and non-clitic occurrences, only the latter being able to appear in situ. There are no clitics for other types of pronouns or nouns, hence the non-available (N/A) mention in the table. For nominal direct objects other than personal pronouns, the default use is in situ and unmarked, hence we counted only the DOM marked occurrences, as our purpose is to see how DOM interacts with CD.

The gist of our findings is that hidden CLLD is more advanced in genuine Romanian than in the translated text, while overt CLLD considerably lags behind in both language registers. CD without DOM is the least represented option, while DOM without CD is very productive. The CD/DOM interaction is emerging more or less at the same pace as overt CLLD, while the topicalization option is not productive, especially with pronouns.

There is no doubt, according to the statistics, that overt CLLD is tightly related to the exploitation of CLLD for the CD/DOM interaction. The interaction arises from the extension of CLLD to DOM with a minimal adjustment, namely, leaving the KP in situ instead of moving it. This amounts to the underlying structure in (14), which involves long distance Agree instead of Spec-head agreement. It is, however, impossible to determine whether this derivational extension begins with the bleaching of K features, or whether the features of K are bleached because the clitic, carrying equivalent values, was made available in the grammar.

A second piece of evidence for the contrast between (13) and (14) comes from the following observation: a DOM-ed KP undergoing overt CLLD may or may not display pe in MR, as in (16a); conversely, in OR pe is almost obligatorily present in this context, as in (16b) (only two exceptions found in the entire corpus of 16^th century texts).

(12)

(16) a. (pe) aceşti prieteni i-am văzut ieri MR DOM these friends them-have seen yesterday

‘These friends I saw (them) yesterday.’

b. pre aceşti boiari i-au băgat în temniţă (DÎ, XVIII, TR, 1599) OR DOM these lors them-have thrown in jail

‘These lords they thrown in jail.’

The obligatory presence of pe in (16b) indicates a transitional phase, when the division of tasks between pe and the clitic was not well established, so both had to be present.

In fact, the contrast in (16) allows us to verify a theoretical prediction arising from the feature checking analysis. Feature checking is, theoretically, possible not only through the direct merge of an item, but also through Spec-head agreement.

Can K[utop] be checked in any other way than the direct merge of pe? We take the optionality of pe in (16a) to illustrate the Spec-head option. More precisely, the difference between +pe and –pe configurations in TopP is captured in (17a) and (17b), respectively.

(17) a. [TopP KP [TP ClP [v/VP <KP> [V [<ClP> <KP> Cl [<KP>...] +pe fronting b. [TopP DP [TP ClP [v/VP <DP> [V [KP <DP> [ <DP> ...]]]]]] –pe fronting

In (17a), we see the movement of the entire KP to Spec, TopP, in which case pe remains in K, as in (14). In (17b), we see only the DP moving cyclically through Spec, KP and further, thus checking the feature of K, so pe is not necessary. This movement, however, depends on the Agr relation between D and the clitic, which ensures that Spec, KP is argumental. When this Agr relation is absent, as in (13), with DOM without CD, then Spec, KP is a non-argumental position, and movement of the DP to Spec, KP triggers Criterial freezing (Rizzi 2010), so no further movement to Spec, TopP is possible. From this perspective, (16b) signals a transitional phase when D agrees with the clitic, but the A/A’ status of Spec, KP is ambiguous, so the safest option (i.e., pe merging) is preferred.

To summarize the formal analysis, we can capture the interaction between CD and DOM at two levels: (i) the internal structure of the nominal DO, as in (18);

and (ii) the movement within the clause, as in (19).

(18) [KP K-pe[utop] [DP D[δ]/φ] [NP...]]]]

(19) [TopP Topλ [TP ClP [VP KPλ [V [<ClP> <KP> Cl [<KP> pe-DP....]]]]]]

In (18), K has an underspecified and uninterpretable [top] feature, and transfers both δ and φ sets to D, which yields the typological option for Category III in (8), namely Spanish and other Romance languages. In (19), V selects a Clitic Phrase as its direct object, and the clitic head further selects KP, along the lines in Delfitto (2002). KP moves cyclically to Spec, VP under the th-role probe, where it receives Accusative Case. The cyclical movement of KP through Spec, ClP results

(13)

in the φ-agreement with the clitic due to the Spec-head relation. The vacated ClP moves to the TP domain, in response to the δ probe on T (C to T transfer) and thus opens the λ-abstract in TopP. This analysis correctly predicts that weakend pe cannot occur in KPs that do not fall under CD, so the interaction between CD and DOM becomes obligatory in MR.

Basically, the analysis proposed here amounts to a refinement of Delfitto’s patterns as follows. For Delfitto, CLLD involves two patterns: one in which we see only the clitic, because the KP is null (hidden CLLD; e.g., proj lj-am chemat); and one in which we see not only the clitic but also the lexical KP fronted to Spec, TopP (overt CLLD; e.g., pe Ionj lj-am chemat). Since movement occurs in both patterns, this is taken as the trigger for Spec, TopP saturation, with no further implication of featural make-up. However, (19) shows a CLLD variation, where movement is not sufficient to justify the saturation of Spec, TopP, since KP remains in its th-position. This is the point where Agr-δ is needed to justify the long distance Agree between Top and KP, by which Spec, TopP is saturated. The in situ option for KP does not arise with typical CLLD, because the KP object does not have the discourse [top/δ] features, so there is no trigger for an long distance agreement relation with Top/clitic as wee see with CD. Constructions as in (16a) show that DOM as in (19) does not prevent KP from moving to Spec, TopP, but that depends on the expressivity option at clause level, in addition to the Agr-δ/φ chain in place.

The non-trivial consequence of the representations in (17) to (19) is that CD without DOM, further illustrated in (20), must also be a sub-case of CD/DOM interaction.

(20) dentr-o maje de aur să-l faci acesta (PO, 263) from-a ton of gold SUBJ-it make this

‘Make this from a ton of gold.’

At the first sight, (20) should not be possible under Delfitto’s (2002) predictions: the λ-abstract in Top must be checked either by the movement of the KP to Spec, TopP, which does not take place; or by long distance Agr with a D[δ]/φ], the latter implying DOM, which is not visible in (20). One may say that the unproductivity of this construction in OR follows from this untypical configuration. However, CD without DOM as in (20) is productive in other languages, including Aromanian. The analysis we proposed in this paper can cover such constructions as being a variation on (17b): while in (17b) DP moves through Spec, KP, in (20) it remains in Spec, KP. In other words, there is DOM, since K[top] probes DP, and an Agr relation is established between the clitic and D. The choice between pe or DP as checking devices for K[top] depends on cross- linguistic preferences, which may be related to other semantic variation in the option for DOM in general (this is subject to further research).

(14)

To conclude this section, we proposed that OR shows the use of pe-DOM at two stages of reanalysis: one where pe spells out a complet set of discourse features on K (i.e., contrastive topic and δ) and DOM has a foregrounding effect; and one in which pe spells out an underspecified [top] feature on K, while δ is transferred with φ to D, in which case DOM has a backgrounding effect. The former refutes CD, since pe is sufficient to check the features of K, whereas the latter needs CD, since pe checks [top] but not δ on D, and brings no valuation to these features. The checking/valuation task is taken over by the clitic. This analysis correctly allows for the semantic peculiarities discussed in the literature; for example, the semantic classes of nouns are less restricted with the foregrounding pe-DOM than with the CD/DOM, and featural shifting towards animacy and deixis is increased after the bleaching of the discourse features on K.⁴

5. PARAMETRIC CHANGES

In terms of parametric changes, Miyagawa’s (2010) typology in (8) indicates that OR displays a transitional period from a grammar of Category II, like English, with δ at C and K (i.e., fronting through topicalization in CP; contrastive topic for DOM), to a grammar of Category III, where δ is transferred to T and D (i.e., fronting through CLLD in CP; familiar reading for DOM). This parametric shift is not peculiar to Romanian but affected other Romance languages as well. The actual implementation of this shift shows cross-linguistic variations within the Romance group, according to the morphological properties of each language. For example, where CD/DOM is concerned, Spanish accommodates the parametric shift within the analytical Dative paradigm, whereas OR switches to the analytical Accusative.

It is, thus, predictable that CD/DOM exhibits some cross-linguistic micro differences in both syntactic and semantic areas (Irimia 2017).

The parametric shift is completed in MR, where operator topicalization is lost, and hidden CLLD is the only way of mapping the direct object through a personal pronoun for neutral readings. When it comes to DOM, CD/DOM is generalized, with very few traces of foregrounding DOM seen mostly in the grammar of old age speakers. Only bare quantifiers are used in standard MR with foregrounding DOM –see (21a, b). However, this is subject to inter-language variation as well, since CD-ed bare quantifiers do appear in both OR and non- standard MR, as shown in (21c, d). Arguably, the shift is still to be completed in this respect.

4 Heusinger and Onea Gáspár (2008) propose the concept of anchoring specificity to capture the interpretive effects of DOM: DOM without CD is locally anchored, whereas CD/DOM is discourse anchored. According to the analysis proposed here, this difference is read off the syntactic configuration, which involves narrow scope within KP for the foregrounding DOM, but wide scope within CP for CD.

(15)

(21) a. (*L)-am chemat pe cineva.

him-have called DOM somebody ‘I/We have called somebody.’

b. Pe cine (*l)-ai chemat ? DOM who him-have called ‘Who did you call ?’

c. tremeate-l pre cine ţi-e voia (PO, 154) OR send-him DOM who you-is will-the

‘Send the one you want.’

d. Pe cine l-a propus ca ministru ?⁵ MR DOM who him-have proposed as Minister

‘Who did he propose as a Minister?’

A final note must be made about Miyagawa’s typology: his Agr-δ feature covers the functional relations of all the discourse pragmatic features, which means both topic and contrastive focus types. Here we only discussed the topic type, since it concerns DOM, but focus was equally affected by the parametric shift. For example, in OR, contrastive focus involved operator driven movement of either XP to Spec,FocusP or of V-to-Focus (Alboiu et al. 2015). However, in MR, V-to- Focus is lost and KP-to-Spec, FocusP is subject to CLLD on a par with movement to Spec,TopP (e.g., Pe el *(l)-am chemat, nu pe ea). An investigation of contrastive focus with CLLD is beyond the scope of this paper, but it needed mentioning in order to confirm the consistency of the parametric shift.

6. CONCLUSIONS

This paper started with a double-fold question: when and why did CD interact with DOM in Romanian? The answer to the when question is 16^th century for the incipient stage with fast increase by the 18^th century and further, as CLLD spreads. The answer to the why question is the following: DOM involves K with a topic feature, and this topic feature underwent bleaching, leading to structural changes. These changes are signalled by the reanalysis of pe and the changes in the readings of DOM, from contrastive to familiar topic. We argued that CD inched into DOM derivations in order to supplement the checking activity of pe.

Theoretically, the interaction between CD and DOM was treated as a type of CLLD. Accordingly, the emergence of CD/DOM was contingent on the productivity of CLLD in the language, which was shown here (for the first time) to gain foot against non-CLLD options.

Typologically, this analysis allowed us to conclude that a parametric shift took place in OR, whereby the feature responsible for the generation of clitic

5 Source: adevarul.ro/news/politica/ 1_57bba41e5ab6550cb8020d9a/index.html

(16)

constructions (i.e., Agr-δ) was transferred from the discourse to the inflectional domain in both clause and nominal phrases (i.e., from C to T and from K to D, respectively). Thus, the parametric shift does not concern DOM per se, but the emergence of clitic constructions in general.

REFERENCES

Aboh, E., 2004, “Topic and focus within D”, Linguistics in the Netherlands, 21, 1–12.

Aboh, E. et al., 2010, “DP-internal information structure: some introductory remarks”, Lingua, 120, 782–801.

Alboiu, G. et al., 2015, “Discourse driven V-to-Focus in Early Modern Romanian”, Natural Language and Linguistic Theory, 33, 4, 1057–1088.

Avram, L., R. Zafiu, 2017, “Semantics hierarchies in the diachronic evolution of differential object marking in Romanian”, in: A. Dragomirescu et al. (eds), Sintaxa ca mod de a fi. Omagiu doamnei profesoare Gabriela Pană Dindelegan, la aniversare, Bucureşti, Editura Universităţii din Bucureşti, 29–43.

Bossong, G., 1991, “Differential object marking in Romance and beyond”, in: D. Wanner, D. Kibbee (eds), New Analyses in Romance Linguistics, Selected Papers from the XVIII Linguistic Symposium on Romance Languages 1988, Amsterdam, Benjamins, 142–170.

Chiriacescu, S., K. von Heusinger, 2010, “Discourse Prominence and pe-Marking in Romanian”, International Review of Pragmatics, 2, 2, 298–332.

Chomsky, N., 2008, “On Phases”, in: Freidin, R. et al. (eds), Foundational Issues in Linguistic Theory: Essays in Honor of Jean-Roget Vergnaud, Cambridge, MA, MIT Press, 133–167.

Comrie, B., 1989, Language universals and linguistic typology, 2^nd edition, Chicago, University of Chicago.

Delfitto, D., 2002, “On the semantics of pronominal clitics and some of its consequences”, Catalan Journal of Linguistics, 1, 29–57.

Dobrovie-Sorin, C. 1990, “Clitic doubling, wh-movement and quantification in Romanian”, Linguistic Inquiry, 21, 351–397.

Farkas D., K. von Heusinger, 2003, “Stability of reference and object marking in Romanian”, paper presented at the Workshop on direct reference and specificity, ESSLLI, Vienna.

Frascarelli, M., R. Hinterhölzl, 2007, “Types of Topics in German and Italian”, in: K. Winkler, S. Schwabe (eds.), On Information Structure, Meaning and Form, Amsterdam, John Benjamins, 87–116.

Gierling, D., 1997, “Clitic doubling, specificity and focus in Romanian”, in: J. Black, V. Motapanyane (eds), Clitics, Pronouns and Movement, Amsterdam, John Benjamins, 63–86.

Giusti, G., 2006, “Parallels in Clausal and Nominal Pheriphery”, in: M. Frascarelli (a cura di), Phases of Interpretation. Studies in Generative Grammar, vol. 91, Berlin, Mouton de Gruyter, 163–186.

Giusti, G., 2012, “On Force and Case, Fin and Num”, in: V. Bianchi, C. Chesi (eds), Enjoy linguistics!, Siena, Centro Interdipartimentali di Studi Cognitivi sul Linguaggio, 205–217.

Haegeman, L., 2004, “DP-Periphery and Clausal Periphery: Possessor Doubling in West Flemish”, in:

D. Adger, C. de Cat, G. Tsoulas (eds), Peripheries. Syntactic Edges and Their Effects, Dordrecht, Kluwer, 211–240.

von Heusinger, K., E. Onea Gáspár, 2008, “Triggering and blocking effects in the diachronic development of DOM in Romanian”, Probus, 2, 67–110.

Hill, V., 2013, “The Direct Object-marker in Romanian: a historical perspective”, Journal of Australian Linguistics, 33, 2, 140–151.

Hill, V., L. Tasmowski, 2008, “Romanian Clitic Doubling: A view from pragmatics-semantics and diachrony”, in: D. Kalluli, L. Tasmowski (eds), Clitic Doubling in the Balkan Languages, Amsterdam, John Benjamins, 135–164.

(17)

Irimia, M., 2017, “Licensing differential objects: some remarks on Romanian”, Ms.

Kayne, R., 1975, French Syntax, Cambridge, MA, MIT Press.

Lambrecht, K., 1994, Information structure and sentence form: Topic, focus and the mental representation of discourse referents, Cambridge Studies in Linguistics 71, Cambridge, Cambridge University Press.

Lee, C., 2003, “Contrastive Topic and/or Contrastive Focus”, in: McClure, B. (ed.), Japanese/Korean Linguistics, Stanford: CSLI, 352–364.

Mardale, A., 2015, “Differential Object Marking in the First Original Romanian texts”, in: V. Hill (ed.), Formal Approaches to DPs in Old Romanian, Brill, Leiden-Boston, 200–245.

Massey, V., 1991, Compositionality and Constituency in Albanian, PhD dissertation, University of North Carolina at Chapel Hill.

Mišeska -Tomić, O., 2006, Balkan Sprachbund Morpho-syntactic Features, Dordrecht, Springer.

Miyagawa, S., 2010, Why Agree? Why Move? Unifying Agreement-based and Discourse Configurational Languages, Cambridge, Mass, MIT Press.

Pană Dindelegan, G., 2014, “Variaţie în construcţia verbului în româna veche”, in: R. Zafiu et al.

(eds), Limba română: Diacronie şi sincronie în studiul limbii române, Bucureşti, Editura Universităţii din Bucureşti, 155–176.

Pesetsky, D., E. Torrego, 2007, “The Syntax of Valuation and the Interpretability of Features”, in:

S. Karimi, V. Samiian, W. Wilkins (eds), Phrasal and Clausal Architecture: Syntactic Derivation and Interpretation, Amsterdam, John Benjamins, 262–294.

Ritter, E., 1995, “On the syntactic categories of pronouns and agreement”, Natural Language and Linguistic Theory, 13, 3, 405–443.

Rizzi, L., 2010, “On some properties of criterial freezing”, in: P. Panagiotidis (ed.), “The Complementizer Phrase: subjects and operators”, Oxford, Oxford University Press, DOI: http://dx.doi.org/10.1093/acprof:oso/9780199584352.003.0002 (Accessed Jan 2017).

Szabolcsi, A., 1994, “The Noun Phrase”, in: F. Kiefer, K. E. Kiss (eds), The Syntactic Structure of Hungarian, New York, Academic Press, 179–274.

Suñer, M., 1988, “The role of Agreement in clitic-doubling constructions”, Natural Language and Linguistic Theory, 6, 391–434.

Tigău, A., 2015, “Romanian Differential Object Marking and Presupposition”, Nouveaux cahiers de linguistique française, 32, 215–227.

Wiltschko, M., 2014, The universal structure of categories, Cambridge, Cambridge University Press.

(18)